The onward march of algorithmic justice: a UK warning

You don’t want to get too deeply into the intricacies of UK Brexit immigration policy if you can help it. So, you might want to read the latest publication, Quick and Easy Justice, from the Public Law Project by Joe Tomlinson (praised in June for his general pamphlet Justice and the Digital State) as a continuation of his argument rather than a specific study of the arcane processes governing EU citizens seeking to regularise their residence in the UK. All you really need to know is that around 3.5 million people – and probably more (no one is quite sure) – must have their applications considered for residence (provided that Brexit actually happens) and the government wants to keep the cost down by automating the process as much as it can. Joe Tomlinson outlines some of the likely problems. Recent publications from others suggest that he is correct to predict that automation and algorithms raise issues.

The process designed for claims to ‘settled’ or ‘pre-settled’ status is largely online. The ‘streamlined application generally relies on two platforms: an app downloadable on a mobile phone or tablet, and an online form filled on the UK Government’s website. Those who fall within the personal scope of the Scheme must submit information on both of these two platforms which evidences three broad categories: identity, residence, and suitability …’ You eventually get to scan in your photo and your evidence. 

‘The automated part of the application process will use an algorithm to check HMRC [Her Majesty’s Revenue and Customs] and DWP [Department of Work and Pensions] data for proof of residency. Specifically, three fields of data—an applicant’s name, date of birth, and national insurance number — are sent automatically to the DWP and HMRC. Once this information has been received by those two Departments, it is transferred to a “Citizen Matching Layer,” which identifies the applicant and searches the respective Departmental databases for details about the matched applicant. The information is then relayed back to the Home Office and transferred to its “business logic” – an algorithm which is yet to be disclosed publicly–which processes the information to establish the period of continuous residence in the UK.’ The result is a pass, a partial pass or a fail, the decision itself technically taken by an official. 

The system has been trialled, but only on fairly stable and tame audiences. It has not yet been in operation long enough to be properly examined. But, a recent blog from Kuba Jablonowski on the London School of Economics Brexit blog is worth consulting on suspicions of what is actually happening. You have to remember that the Home Office, the department in charge of immigration, is a byword for a particularly unattractive mix of incompetence, delay and hostility. It is still reeling from the ‘Windrush scandal’, where Caribbean immigrants from the 1960s were wrongly threatened with deportation because, at least in part, the Home Office itself had destroyed their records. So, there is a general cynicism among immigration advisers about whether the Home Office can manage any project like this efficiently and fairly. 

The automated part of the process involves the DWP which, alas, has a similarly poor reputation which might introduce an unintended bias.  HMRC is better but there is wrinkle in the system: ‘data from working tax credit, child tax credit, and child benefit records, all managed by HMRC, is not being shared. As it is more likely that women are in receipt of these benefits, there is a risk that the exclusion of this data means women are at a greater risk of not passing the automated check.’ So, there is a clear potential gender bias in the system.

The PLP paper is an analysis of potential problems. Mr Jablonowski of the LSE Brexit project has some data. He is worried that volumes of applications are down after an early rush: ‘The volumes of applications concluded in June and July were down versus April and May, the first full two months of the Scheme’s operation. What is concerning too is the application backlog. While it is now below 90,000 open cases, it was not cleared as application volumes slowed down and this again raises questions.’ The press has covered a number of  high profile decisions that appear just wrong. A number of reports have analysed the problems and made recommendations for improvements – for example one from British Future.

A particular problem seems to be emerging. Not only are some good applications being refused, an increasing number of applicants are receiving pre-settled status instead of settled status despite having lived in the UK for well over five years (which should entitle them to a right to settle). ‘Their stories fit into a worrying overall trend …, ‘ writes Mr Jablonowski, ‘of the rise in decisions granting pre-settled status. After the first full month of the Scheme’s operation in April, 66% of applicants were getting settled status. In July, this ratio dropped down to just 56%.’   The percentage getting pre-settled status rose to 42 per cent in August, having been 32 in the testing phase. It looks like official decision-makers are hedging their bets; deciding cases on effectively an interim basis; and figuring that everything will be sorted out in the future.

Clearly, the advice and legal sector in the UK has a major role in connoting to identify problems with this new scheme. But you can already see some general lessons. Automated checking between government departments is not, of itself, wrong. It should represent desirable collaboration. But, if the DWP is not reliable and the tax information is partial, then they cannot be over-relied upon. The great danger with algorithms and automation of this kind is that it will encourage an under-investment in schemes which needs high grade decision-makers (after all residence is a pretty important human right) and major outreach (since there is a cut off date for applicants). British Future estimates that ‘It is likely that about 70% of EU citizens in the UK will have correct information about the EU Settlement Scheme and will be able to submit their applications with ease. But research  suggests that about 30% of EU citizens risk being left out, for reasons which relate both to people’s individual circumstances and the way the system operates.’ There must be rapid and effective ways of challenging decisions, with a constant eye to the inadequacy of departments like the DWP or the partial nature of the information from the HMRC.

There is a further point. The social security reforms of the early 1980s briefly introduced into legislation the notion of a post which became known as the ‘adjudication officer’, an attempt to recognise the need for more independence in departmental decision-making. There was a Chief Adjudication Officer. Perhaps we need to revive this idea for all decision-makers in government departments, particularly as more use of algorithms and automation is made in decision-making.  

Leave a Reply